远焦区聚焦性能分析与改善  

Performance analysis and improvement on acutely defocused scenes

作　　者：孙明磊[1] 石晶欣[1] 董代[2] 宗光华[1] 

机构地区：[1]北京航空航天大学机械工程及自动化学院,北京100191 [2]北京航天控制仪器研究所,北京100039

出　　处：《北京航空航天大学学报》2008年第9期1032-1036,共5页Journal of Beijing University of Aeronautics and Astronautics

基　　金：国家863计划资助项目(2004AA404260);国家“十五”“211”学科建设资助项目

摘　　要：传统聚焦测度函数在远焦区对图像高频能量解析性能较差;表现为聚焦曲线在远焦区呈现非单调性.利用傅里叶变换方法对图像频谱进行分析:在远焦区,聚焦图像序列的直流分量表现出强烈的波动性,并且直流分量在图像频谱中占据绝对比例,再加上传统聚焦函数滤除图像直流分量的能力较弱,导致误聚焦概率高.考虑到频域聚焦函数计算量大,提出基于时域的两个聚焦函数.实验结果表明:此二函数具有强的直流分量滤除能力,能有效改善聚焦曲线在远焦区的表现,同时具有高的聚焦分辨率.远焦区聚焦性能的改善对于提高显微视觉操作自动化的可靠性尤为重要.Some classic focusing measures such as gradient function, Brenner function and entropy function, et al, show poor performance in computing high frequency energy in acutely defocused images. Their focusing curves away from the focus show serious nonmonotonicity. The Fourier transform method for image series spectrum analysis was conducted. Three reasons for poor performance in far away focus region were proposed. Firstly, defocused image sequences show strong volatility in power of DC （direct current） component. Secondly, DC component of the image spectrum occupies an overwhelming proportion. Thirdly, the traditional measures have weak ability in filtering DC component of image. According to this, two spatial focusing measures were presented. Experiment results show that performance of focusing curves of acutely defocused images was significantly improved, and high focus resolution can be achieved. This method is particularly important in enhancing reliability of automatic micromanipulation based on microscopic vision.

关 键 词：视觉 聚焦 图像 傅里叶变换 微操作 

分 类 号：TP242.62[自动化与计算机技术—检测技术与自动化装置]